INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 conduc

-0.74

æľ

-0.71

endas

-0.70

 Bullets

-0.67

URI

-0.64

anwhile

-0.63

éĩ

-0.62

 Ambro

-0.62

Cu

-0.61

İ

-0.61

POSITIVE LOGITS

 ident

0.72

bent

0.66

employed

0.65

icularly

0.64

aird

0.64

 Gust

0.64

 Hunt

0.63

 WATCHED

0.62

lihood

0.61

folk

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.